When I read a physics textbook about a P-N junction, it will tell me that you connect a P-doped semiconductor to an N-doped semiconductor, and you form a depletion region.
From what I know about silicon wafer manufacturing, wafer producers will make an ingot using some silicon melt that has some doping built in. That ingot is then cut into many wafers.
If a wafer is either P-doped or N-doped, how do you form a P-N junction? Do we connect two oppositely doped wafers?
Bonding can be done, https://en.wikipedia.org/wiki/Direct_bonding but it's not easy. If you think about having to achieve atomic flatness and freedom from contamination, over a massive area... yeah. That it's doable at all is truly testament to the incredible precision used in these processes.
More often, either extra doping is introduced (by chemical deposition and diffusion i.e. physically introducing a dopant and baking it in, or more often in modern processes AFAIK, directly by ion implantation), or material is added on top by a chemical deposition process (epitaxy), which can contain dopant gasses thus giving a high purity N/i/P (as the case may be) layer. [i = intrinsic i.e. non-doped]
The fact that this is ultimately a printing process, is very valuable even to simple devices like diodes, as edge effects dominate their behavior at extremes of operating range, particularly surge immunity and leakage current. Sawing up a flat sandwich of P/N leaves a bare edge around the junction, contaminated by the sawing process, and subject to surface states; instead doping the surface allows the junction to curl up towards the surface, where it can be carefully controlled as part of the printing process (e.g. adding guard rings to control surface states, electric field profile, etc.).
These more subtle effects might not be important to solar cells (I forget if they are cleaved/sawed, or printed as above?), but essentially all commercial electronic semiconductors are made this way.
You implant dopant ions using an ion accelerator. The ion energy determines the likely implantation depth.
Then the wafer is heat treated, to incorporate the dopants into the Si lattice.
To laterally define the dopant region, a thick mask is prepared on the wafer prior to implantation.
That way you can create an n doped region inside of a p doped wafer, or p doped regions inside previously n doped regions etc. Look at a schematic crosssection of an NPN transistor for an example.
It doesn't have to be a wafer you can be done mechanically. Manufactured dials usually start out with a substrate of silicon that is doped or at dope layers on top of the silicon (or other semiconductor materials)
The cool thing is in the early days of electronics they didn't have wafers and so they used crystals or whatever they had around to mechanically form and junctions or other junctions that had a diode like capabilities.
You could do this at home if you had some kind of semiconductor Crystal.
Modern devices rely on epitaxy, both in Silicon and in III-V materials (GaAs, InP, InAs, GaSb, including GaN, SiC). As you listed, ingots are cut into substrates with certain doping type. Then, to form wafers, there is epitaxial growth of layers on top of substrate to get wafer, with semiconductor device structure. In late XX century, the great breakthrough were epitaxial growth techniques LPE, MBE, MOCVD, where you could form complex bandgap and doping structures of transistor, lasers, detectors, modulators. Doping of grown layers could be engineered in-situ during growth using dopants like Boron, Arsenic..
